cost-effective ert technique for oil-in-water measurement

Cost-Effective ERT Technique forOil-in-Water Measurement for Offshore

Hydrocyclone Installations

Petar Durdevic ∗ Leif Hansen ∗ Christian Mai ∗

Simon Pedersen ∗ Zhenyu Yang ∗

∗ Institute of Energy Technology, Aalborg University, Esbjerg Campus,Niels Bohrs Vej 8, 6700 Esbjerg, Denmark, e-mail: [email protected], +45

31 75 13 20

Abstract: The goal of this paper is to introduce and design a cost-effective Oil-in-Water (OiW)measuring instrument, which will be investigated for its value in increasing the efficiency of adeoiling hydrocyclone. The technique investigated is based on Electrical Resistivity Tomography(ERT), which basic principle is to measure the resistivity of substances from multiple electrodesand from these measurements create a 2-D image of the oil and gas component in the water.This technique requires the measured components to have different electrical resistances, suchas seawater which has a lower electrical resistance than hydrocarbon oil and gas. This workinvolves construction of a pilot plant, for testing the feasibility of ERT for OiW measurements,and further exploring if this measured signal can be applied as a reliable feedback signal inoptimization of the hydrocyclone’s efficiency. Different algorithms for creating 2-D images andthe feasibility of estimating OiW concentrations are studied and evaluated. From both steadystate and continuous laminate flow perspectives, with respect to the objective which is to usethis measurement for feedback control purposes.

Keywords: Electrical Resistivity Tomography, Water Treatment, Process Control, Oil inWater, Offshore.

NOMENCLATURE

Symbol Description Unit

Qo Hydrocyclone overflow flow m3/hQi Hydrocyclone inlet flow m3/hPi Hydrocyclone inlet pressure kPaPo Hydrocyclone overflow pressure kPaPu Hydrocyclone underflow pressure kPaCu Concentration of oil in the underflow mg/LCi Concentration of oil in the inlet mg/Lε Hydrocyclone efficiency %iαβ Measured current Avαβ Measured voltage drop VGαβ Measured line conductance Sgαβ Line conductance pr distance S ·mvβhigh

Passive electrode high side voltage V

vβlow

Passive electrode low side voltage Vvαhigh Active electrode high side voltage V

vαlow Active electrode low side voltage Vrm Measurement resistance value Ωfframe 2-D Frame rate of measurement Hzfs Sampling rate Hznc Channel count −aαβ Coefficient of electrode-electrode line −bαβ Coefficient of electrode-electrode line −cαβ Coefficient of electrode-electrode line −lαβ Distance between electrodes mα, β Active and passive probe designation −S Set of all measurements −

? Supported by the Danish National Advanced Technology Founda-tion through PDPWAC Project (J.nr. 95-2012-3).

wαβ Measurement weight −x Point in 2-D plane cross-section of pipe mx1 First coordinate of electrode-electrode line mx2 Second coordinate of electrode-electrode line mdmin Minimum distance electrode-electrode line m

1. INTRODUCTION

Hydrocyclones are commonly used to separate oil fromwater downstream the three phase separator in the up-stream offshore oil & gas production, this means thatthe oil concentration which they have to handle is lowas some separation has already taken place in the threephase separator. During normal North Sea operation theoil concentration exiting the three phase separator is below5000 parts per million (PPM) or 0.5%, see Kharoua et al.(2010), while the oil concentration in a normally operatinghydrocyclone’s water outlet, should be around 50 PPM to20 PPM. The goal of using hydrocyclones for producedwater treatment in offshore oil & gas industry is to keepthe concentration of hydrocarbon content (mainly oil part)in the treated water below 30 PPM as this is the currentNorth Sea regulation (Miljoestyrelsen (2010)). A typicalhydrocyclone consists of a cylindrical chamber to which atangential inlet inputs a mixture of oil and water. Due tothe curvature of the cylindrical shape, this fluid mixturewill start to flow in a rotating path. The centrifugal forcesthen act on the fluid, and consequently the fluid withthe higher density is pushed furthest out to the cylinderwall. The cylindrical part is connected to a long conicalpipe, where at the end the water exits, this is called theunderflow (Wolbert et al. (1995)). The conical part pressessome of the lighter fluids upwards to the opposite end of

2nd IFAC Workshop on Automatic Control in Offshore Oil and Gas Production,May 27-29, 2015, Florianópolis, Brazil

Copyright © 2015, IFAC 153

the cylindrical chamber, where a pipe of a small diameteris used to let out the oil, and this is called the overflow(Wolbert et al. (1995)).A schematic illustration of a typicalhydrocyclone is shown in figure 1, where the paths of oil(red) and water (blue) are depicted respectively. During

Fig. 1. A cut-trough image of the hydrocyclone, showingthe hypothetical water (blue) and oil (red) paths.

normal offshore operation, the pressures at the inlet andthe two outlets are measured and used to calculate thepressure drops over the inlet and the two outlets, these twopressure drops are then employed to calculate a so-calledPressure Drop Ratio (PDR) as described by equation 1,which is used to control the hydrocyclone. The PDR iscorrelated to the flow split Qo/Qi, which is the volumetricflow rate of the overflow over the volumetric flow-rate ofthe inflow (Husveg et al. (2007)).

PDR =(Pi − Po)(Pi − Pu)

(1)

It has been discovered by Husveg et al. (2007) that theflow split and the PDR can be related through a linearapproximation. By adjusting the PDR and thus the flowsplit the controller will allow a certain amount of oil to passthrough the overflow and thus ensure an efficient operationof the hydrocyclone. The PDR is linearly approximated tobe proportional to the flow split, and the flow split canbe further related to the efficiency which is illustrated infigure 2 (Husveg et al. (2007)). The hydrocyclone efficiencyis usually defined as a percentage of the concentration ofoil in the underflow Cu over the concentration of oil inthe inlet Ci, resulting in the following equation ε = 1 −Cu/Ci, see Young and Wakley (1994). This type of PDR-

Fig. 2. Hydrocyclone efficiency related to the flow split, theresult is based on an empirical analysis of a typicaloffshore liner (Husveg et al. (2007)).

based feedback control loop, which is the most commonlyused in the offshore E&P, does not measure the OiWconcentration and thus has no direct control regarding thequality of the treated water exiting the hydrocyclone. Thedevelopment of the PDR control is thus done based onplenty empirical knowledge of the system, and the refer-ence value given to the closed-loop system varies for differ-ent systems and is typicality kept at around 2-3 (Husveget al. (2007)). One of the effective techniques commonly

applied for measuring multiphase flow in the offshore in-dustry is to separate different phases and then measureeach phase individually using different single phase flowmeasurement (Ismail et al. (2005)), (Thorn et al. (1999)).However this kind of approach is not a directly applicablesolution for feedback control of hydrocyclones, with re-spect to the fact that different measurements could variateand lead to some unpredictable time-delay, i.e. from thetime the mixture enters the separation process until itis separated and measured. This type of measurement ismore sensibly used, for example, for logic-level supervisorycontrol or offshore evaluation. This work aims at usingElectrical Resistivity Tomography (ERT) for measuringthe OiW concentration at the hydrocyclone’s inlet andoutlets. Compared to other relevant techniques, such as γ-ray tomography, the proposed ERT technique has no useof radiation, as well as to its non-invasive and structurallyrobust and cost effective characteristics. Other alternativesolutions could be, such as the Advanced Sensors EX−100(Advanced Sensors (2014)), which uses Laser Induced UVFluorescence to measure the amount of hydrocarbons inwater. This instrument can measure in various ranges,from 0−10 PPB, to 20.000 PPM, which can be extremelysensitive and precise. With regard to real-time purposethis instrument can have a sampling-rate of 1sec, whichcould be applicable for feedback control purpose. Howeverthis type of sensoring instrument can be very expensivein a range of 30,000-40,000 USD. From our experiencethis equipment is difficult to calibrate in ranges of around5−50 PPM, mainly due to the coalescence of oil in water.Normally the test samples need to be thoroughly mixedbeforehand. Other available commercial instruments, likethe Agar Corporation MPFM 50 Multiphase flow meterwhich could be found at Agar Corporation (2014), andthe Jorin VIPA which could be found at Jorin (2014).Nevertheless, all of these mentioned products are quiteexpensive. Thereby instead of concentrating on the usageof these sophisticated and expensive equipments, we choseto concentrate on some cost-effective electric tomographyfor OiW measurement, for instance, the constructed to-mography sensor used in this paper costs less than 20 USD.The potential economic benefits of using cost-effectiveERT for OiW concentration measurement is quite obvious.In addition, by analyzing the time-dependent tomographyresults, dynamic changes in the flow-rate and flow-regimecan also be detected, which allows for a wide range ofapplications, such as slug-detection, separator and pipelinemonitoring (Dong et al. (2001)). The technique is alsoscalable for both the pipe size and number of electrodes tosuit the fidelity/resolution requirements of each individualapplication. In general, more electrodes would give betterresolution but with slightly increased cost and computa-tion requirements.

2. TOMOGRAPHY TECHNIQUES

Tomography is a non-invasive technique for visualizingsome information over a cross-sectional segment of a pipe,which could be filled with two components with differentelectrical characteristics. Measurements are usually takenfrom several points around the pipe and the test results arethen used to create a image of the segment. Tomographydata can be retrieved based on different measurement tech-niques, such as: electrical, radioactive, optical, microwave,

IFAC Oilfield 2015May 27-29, 2015


ultrasonic and magnetic resonance, where the last is wellknown from MRI scanning used for medical imaging (Is-mail et al. (2005)). Our work will concentrate on electricaltomography as this could be the cheapest and most simpletechnique to implement, and it is also structurally robustwhich is the primary objective for the offshore instruments.The application of ERT on solid-liquid hydrocyclones ap-plied in the clay industry has been reported in Williamset al. (1999). A test setup is introduced with electrodes inseveral segments of the hydrocyclone such that differentperformances at different segments inside the hydrocyclonecan be possibly observed. This work discovered that itis possible to determine if an air-core is formed insidethe hydrocyclone as well as which type of underflow dis-charge is occurring. However the approaches and methodsproposed in Williams et al. (1999) cannot be directlyapplied for de-oiling/liquid-liquid hydrocyclones, due tothe dramatically different structural and operational char-acteristics of liquid-liquid hydrocyclones from solid-liquidhydrocyclones. For instance, within one of actual cases weexperienced, the offshore liners often receive the flow fromthe separator with a maximal amount of 5% which is equalto 0.5 % mass fraction of oil in water (Kharoua et al.(2010)), while according to Williams et al. (1999), theirclay hydrocyclone units receive up to 15 % mass fractionof solids. Also the pilot plant in Williams et al. (1999) in-volves installation of electrodes on the hydrocyclone wall.In the offshore industry this is not possible, with respectto the very high safety specifications in the constructionsof equipment. We will therefore concentrate on installingthe electrodes upstream the inlet and downstream theunderflow of the hydrocyclone. Dong et al. (2003) inves-tigated ERT for measuring two phase flow in pipes interms of water and air, their results showed the potentialcapability of ERT instrument of measuring different typesof flow regimes. They successfully distinguished betweenBubble, Slug, Multi-Bubble and Annular flow. Two andthree dimensional images were successfully constructed ina real-time manner. We believe that once techniques canbe successfully used for gas with water, they can also bepotentially used for oil with water with respect to thefact that both air and oil have very low conductivities.Hydrocarbon oil has a conductivity less than 10pS/m(Michael Lindner (2014)) and a gas mixture such as airhas a conductivity as low as 2.95fS/m with high aerosolconcentration (Pawar et al. (2009)). These two conduc-tivities both have much lower values than water does, forinstance the water with a salinity of 20g/kg and 20Chas a conductivity of 2.901S/m according to NPL (2014).Therefore these techniques could be extrapolated and usedto examine OiW concentrations in evaluating deoiling sys-tem’s performance. Ismail et al. (2005) pointed out somepotential drawback of using electric tomography methods,i.e. their inability to send electrons in direct paths, which isdifferent from γ-ray where the rays can be sent in a directpath from the source. Thereby using ERT, there is somerisk that the electrical current from one electrode to theother through the medium may travel around obstacles ofhigh resistivity, by following the path of least resistance.Apart from the direct path which the γ-ray tomographycan handle, γ-ray tomography is also able to make fastmeasurements, Thorn et al. (1999) mentioned that theirsystem developed at the University of Bergen can handle

several hundred frames per second if a sufficient computingpower is available. Johansen et al. (1996) mentioned thesafety issue of using γ-ray tomography, though with theirsetup the radiation is less than 0.1µSvh−1 at 1 metersdistance, with is far below the recommended maximumdose of 7.5µSvh−1. To achieve this low radiation, thesetup described in Johansen et al. (1996) requires carefullyconstruction using thick steel plates to shield the radiation.Although the γ-ray method has some advantages over theelectric tomography, as shown in figure 3, it won’t be themain focus in this work, mainly due to the complexity ofdesign and the high price of it.

Fig. 3. Compassion between ECT and γ-ray tomography,the test benchmark is illustrated as the middle image.The γ-ray tomography can be improved to show thewater phase (Johansen et al. (1996)).

3. TEST SETUP

The ERT method is, from the manufacturing perspective,cheap and easy to construct, however it requires directelectrical connection to the pipe contents thereby thismethod requires the use of somewhat specific materials.The resolution of the method depends on the permissiblenumber of electrodes. The test setup is constructed toenable the liquid to be stationary and to be flowing subjectto a pumping system. This setup consists of several parts:the ERT pipe which is constructed out of polymethylmethacrylate (PMMA, acrylic glass) with a nominal diam-eter of 50mm and 30cm length; a centrifugal pump; con-necting pipes with an inner diameter of 12mm and a buffertank. A schematic diagram of the test setup is illustratedin figure 4. Water and oil can be added to the system byfilling up the buffer tank. The electrodes have been placed

Fig. 4. Diagram of the test system, containing a control-lable pump, and the ERT unit which is connected toa I/O card. (National instruments PCI-6229)

around the pipe in 360 degrees to cover the entire circle.12 electrodes are placed 30 apart, as illustrated in figure5. The electrodes are threaded stainless steel rods of equallength, which can be adjusted by screwing and whereto themeasurement connections are made using cable clamps.Further mounting improvement could be achieved by usingmachined non-reactive metal electrodes (for example plat-inum, titanium or “Hastelloy”, depending on the processconditions), similar to the design and materials used inmagnetic flow meters, see Emerson (2014). Each electrodeis connected to a digital voltage source through a precision



Fig. 5. Illustration of the ERT sensor configuration

±1% measurement resistor, where each end of the resistoris connected to an ADC channel on the NI PCI-6229 dataacquisition card, as illustrated in figure 6. Conductance

Fig. 6. Probe layout and electrical connection with mea-surement resistors for ERT setup. Red wires representexcitation/measurement connections and green wiresrepresent passive measurement connections

measurement is performed by exciting one electrode at5 volts while all the other electrodes are set to 0 volts,then measuring the voltage on all channels including theexcitation channel. The conductance for one electrode pair,Gαβ , can then be calculated as in equations 2 to 4. α rep-resents the active electrode while β represents the passiveelectrode, such that vαhigh and vαlow is the high add lowvoltages respectably for the measurement resistor at the

active electrode, and vβhigh and vβlow is for the measurementresistor at the passive electrode, see figure 6.Similarly iαβ is the current through the passive electrodeβ wile α is the active electrode, and vαβ is the voltage dropfrom α to β.

iαβ =vβhigh

− vβlow

Rm(2)

vαβ = vαlow − vβhigh

(3)

Gαβ =iαβ

vαβ(4)

After one measurement frame is completed, the sameprocedure will be repeated for all other electrodes. By theend 12 · 11 = 132 (channel) samples are collected. Thesemeasurements can be performed at a maximal samplingrate of fs = 1000[Hz], which is subdivided over thechannel count nc = 16 (due to coding specifics, 4 extrachannels are scanned but not connected to electrodes), sothat each full measurement cycle (frame) can be performed

at a rate as: fframe = fsnc

= 62.5[Hz]. This sampling rate

can be increased by using equipment with a faster analogconverter, if this is required by the application.

4. TOMOGRAM GENERATION ALGORITHMS

In the following, two algorithms, which are based on a2-D plane representing the cross-section of the pipe inwhich all the electrodes are situated, are considered. Theorigin corresponds to the center of the pipe, the x1-axisrepresents the width of the pipe, the x2-axis represents theheight of the pipe. The radius of the pipe is normalized to1. The position of each electrode is represented by a pointin this plane. For each pair of the active and the passiveelectrode (α, β), a standard line equation in the form ofequation 5 is defined for the line trough these points.

aαβ x1 + bαβ x2 + cαβ = 0 (5)

Each of the measured conductances is multiplied with thedistance between the electrodes, lαβ , in order to get aconductance per normalized length unit (radius), gαβ .

gαβ = Gαβ lαβ (6)

For each point in the plane, x, an interpolated conductanceper length unit, g(x), is calculated as a weighted average ofgαβ , which is inspired by Shepard’s method, see Shepard(1968).

g(x) =

∑

(α,β)∈Swαβ(x)gαβ∑

(α,β)∈Swαβ(x)

(7)

S = (α, β)|β 6= α (8)

In this S is the set of all measurements represented by itspair of electrodes, and the weight, wαβ(x) is defined as:

wαβ(x) =1

Di (x, α, β)u(9)

Where the constant u controls how much the influenceof each measurement decreases with distance. Accordingto Shepard (1968) u = 2 is suggested, while Lukaszyk(2004) noted that u > 2 usually is assumed and u > 1is needed for a smooth interpolation function. Di(x, α, β)is one of two different distance quantities denoted eitherDdis(x, α, β) or D Luk(x, α, β), these quantities are whatmakes the two algorithms different in our concern. Bothquantities are related to the minimum distance, d(x, α, β),from a point, x to the line defined for a pair of electrodes(α, β) (see figure 7):

d(x, α, β) =|aαβ x1 + bαβ x2 + cαβ |√

a2αβ

+ b2αβ

(10)

Ddis(x, α, β) is the maximum between d(x, α, β) and a

Fig. 7. Relation between x, α, β and d(x, α, β)

lower bound for the distance dmin, which is used in ordernot to over emphasize the value of a single gα,β for pointson the line defined for the corresponding pair of electrodes(α, β):

Ddis(x, α, β) = max (d(x, α, β), dmin) (11)



The algorithm that applies this quantity is noted as: Min-imum distance algorithm in this paper. D Luk(x, α, β) isan adaptation of Lukaszyk-Karmowski probability metricfor a distance between two random vectors according to Lukaszyk (2004). A new set of axes is defied with theprojection of x on the line for gαβ as origin, the y1-axis ison the line and has positive direction from α to β, the y2-axis is particular on the y1-axis and has positive directionfrom the line to x. The point x is described by Dirac deltadistributions, i.e. an exact value:

fx,1(y1) = δ(y1) (12)

fx,2(y2) = δ(y2 − d(x, α, β)) (13)

While the line is described by an uniform distribution onthe line and a Dirac delta distribution particular on it:

fl,1(y1) =

1

zα+zβy1 if− zα ≤ y1 ≤ zβ

0 otherwise(14)

fl,2(y2) = δ(y2) (15)

Where zα and zβ are the distance from the point’s pro-jection on the line to the active and passive electroderespectively. The definition of D Luk(x, α, β) in equation16 is valid for independent marginal distributions.

D Luk(x, α, β) ≡

2∑i=1

∞∫−∞

∞∫−∞

fx,i(ya)fl,i(yb) dyadyb

p 1p

(16)

Where p defines which p-norm is used to get the norm ofthe vector, as this vector represents a physical distance the2-norm is most commonly used, resulting in equation 17.

D Luk(x, α, β) =

((z2α + z2β

2 · (zα + zβ)

)p+ d(x, α, β)p

) 1p

(17)

The algorithm that applies this quantity is noted as: Lukaszyk-Karmowski algorithm in this paper.

5. TEST PROCEDURE AND CALIBRATION

Two test scenarios are considered: (i) static testing, wherethe ERT-pipe is filled with liquid consisting of a predeter-mined water & oil contents with a mixture ratio of 15/85%,which we name as low water case, and 85/15% which wename as high water case. The static tests are used as thebaseline for exploring the capabilities of the setup; (ii) dy-namic flowing test, where the pipe is subjected to a liquidflow, in order to investigate the dynamic performance ofthe measurement solution. Ideally an oil & water mixturewould have been used, however due to limitations in thepumping system the dynamic flow rate test contains waterand a small gas phase (air) within this early investigationphase, where the air is existing solely due to turbulencein the pipes; no dedicated gas flow was added to this test.In order to calibrate the electrode connections, the pipe isfilled with water, and the electrodes are manually adjustedby screwing until the measured conductivity per length inall paths are the same.

6. TEST RESULTS AND DISCUSSION

The obtained results for stationary low water levels areillustrated in figures 8 and 9, and the high water levelresults are illustrated in figures 10 and 11. By comparingboth concerned algorithms, it is evident that the Lukaszyk-Karmowski algorithm can lead to more smooth results

(in the terms of the gradient), whereas in the minimumdistance approach, the results are affected by the electrodepositions, as compared from figures 8a and 9a. However thesmoothing of the Lukaszyk-Karmowski algorithm resultsin a transient between conductive and non-conductivematerial having a lesser gradient than in the minimumdistance cases, as illustrated from the comparison of figures8b and 9b. This increases the difficulty of determining thephase separation from the 2-D results if a gradient basedapproach is utilized. Another option is to estimate thewater fraction using a simple threshold on the conductancevalues over the set of points χ, as shown in equation 18.

wwater =

∑x∈χ(g(x) > gthres)

nx(18)

Where the threshold gthres is selected from the largesttransition of the minimum distance algorithm in the lowwater case, where the transition is easily determined. Inthese experiments the threshold becomes 1.236 ∗ 10−5.Applying the equation to the results yields the waterfractions listed in table 1. It is observed that the tests

Experiment Low water High water

Minimum distance 8.5 % 87.6 % Lukaszyk-Karmowski 12.1 % 87.7 %

Table 1.

with high concentration of water have worse results thanthe tests with a high concentration of oil/gas, here re-ferring to figures 8a, 8b, 9a, 9b, 10a, 10b, 11a and 11b.We observed that the conductivity transition (high-low),which is illustrated the following 1-D figures 8b, 9b, 10band 11b, is occurring more prominently for the low wa-ter concentration tests than the high water concentrationtests. To illustrate this refer to the gradient of the slopein figures 8b and 9b, where the slope has a high gradientand where it is lower in figure 11b, which represents thehigh water concentration calculated using the minimumdistance algorithm. It has a similar gradient as figures 8band 9b, although the oil water transition is more curvedwhen observing the 2-D figure 10a. A possible reasonfor this behavior is that only a small amount of oil/gasaround the electrodes results in almost 0 conductivitymeasurement, thus the oil/gas is more likely to effect ahigh water concentration than the opposite. This effectis worsened with a lower amount of electrodes, as thisleaves relatively big gaps between electrodes and thus thetransition where oil occurs on the electrodes will be larger,and the results will have a less precise illustration of thephases during high water phase concentrations. For thetest with mixed gas/water flow, illustrated in figure 12 and13 the individual gas bubbles cannot be distinguished fromthe liquid due to the low resolution using either method,which is critical if the gas bubble sizes are to be directlydetermined, however the overall content of the pipe iswell represented as being mostly conductive material. Themean conductance, which is calculated and presented infigures 8a, 8b, 9a, 9b, 10a, 10b, 11a, 11b, 12 and 13, canbe used for determination of the pipe contents. Howeverthis is heavily effected by several factors, including salinityand oil composition, therefore some form of calibrationor secondary measurement will be needed to allow thisvalue to be used for control and monitoring purposes. Aninvestigation of the mean conductance of the low waterlevel in figure 8a (0.35348 · 10−5), the mean conductance



of the high water level in figure 10a (2.1784 ·10−5) and themean conductance of the running water and gas mixturefrom figure 12 (2.3279·10−5) indicate that there is a logicalrelation between the mean conductance and the amountof water in the pipe. This means that it is indeed feasibleto use this technique for measuring the concentration ofwater/oil or water/gas in the pipes.

(a) Minimum distance algorithm, 2D

(b) Minimum distance algorithm, average levels

Fig. 8. Low water level using minimum distance algorithm

(a) Lukaszyk-Karmowski algorithm, 2D

(b) Lukaszyk-Karmowski algorithm, average over width

Fig. 9. Low water level using Lukaszyk-Karmowski algo-rithm

(a) Minimum distance algorithm, 2D plot

(b) Minimum distance algorithm, average levels

Fig. 10. High water level using minimum distance algo-rithm

(a) Lukaszyk-Karmowski algorithm, 2D

(b) Lukaszyk-Karmowski algorithm, average levels

Fig. 11. High water level using Lukaszyk-Karmowski algo-rithm

7. CONCLUSION

In this preliminary study, a cost-effective and non-invasivemethod based on ERT technique for OiW measuring andanalysis has been investigated. The ultimate objective ofthis undergoing investigation aims for a potential cost-effective real-time measurement technique for OiW con-centration at the hydrocyclones inlet and underflow. Totest the feasibility of the proposed method and approach,



Fig. 12. Running water and gas mixture using minimumdistance algorithm, 2-D

Fig. 13. Running water and gas mixture using Lukaszyk-Karmowski algorithm, 2-D

a pilot plant consisting of a 50mm pipe with attached elec-trodes and a pumping circulatory system, is constructed.The oil and water interface in the pipe for static cases canbe clearly observed in the generated 2-D tomograms. Fordynamic flow measurements, where the medium was waterwith a small amount of non-purposed gas mixture, theoverall content of the pipe can still be determined. Troughanalysis of the average conductivity in the mixture, we areable to detect a link between OiW concentration and theaverage of measured conductivities.

Future work will involve testing of the concentration ofoil/gas in water of known concentrations to investigate ifconcentrations can be registered precisely, thus simplifyingthe use of 2-D imaging for feedback control. Anotherextend of our work will be to increase the amount of elec-trodes to allow for higher resolution and thus investigateif this will enable us to visualize small particles in 2-Dimages. Also electric capacitive tomography (ECT) willbe investigated, to compare the merits of this technologywith ERT.

ACKNOWLEDGEMENTS

The authors would like to thank the support from theDanish National Advanced Technology Foundation. Ourthanks also goes to colleagues J.P. Stigkjær, A. Aillos, K.G. Nielsen, M. Haubjerg and P. Molinari from Maersk OilA/S, colleagues P. Sørensen, A. Andreasen, B. Løhndorf,

S. A. Meybodi and J. M. Holm from Ramboll Oil & GasA/S, and H. Enevoldsen from AAU, for many valuablediscussions and technical support.

REFERENCES

Advanced Sensors (2014). Advanced Sensors - Oil in Water Analyz-ers. URL http://www.advancedsensors.co.uk/.

Agar Corporation (2014). Agar Corporation - MPFM 50. URLhttp://www.agarcorp.com/Animation/MPFM/50/MPFM50.html.

Dong, F., Jiang, Z., Qiao, X., and Xu, L. (2003). Application of elec-trical resistance tomography to two-phase pipe flow parametersmeasurement. Flow Measurement and Instrumentation, 14(4-5),183–192.

Dong, F., Liu, X., Deng, X., Xu, L., and Xu, L.a. (2001). Identifica-tion of two-phase flow regimes in horizontal, inclined and verticalpipes. Measurement Science and Technology, 12(8), 1069–1075.

Emerson (2014). Magnetic flowmeter material selection guide.URL http://www2.emersonprocess.com/siteadmincenter/PM%

20Rosemount%20Documents/00816-0100-3033.pdf.Husveg, T., Rambeau, O., Drengstig, T., and Bilstad, T. (2007).

Performance of a deoiling hydrocyclone during variable flow rates.Minerals Engineering, 20(4), 368–379.

Ismail, I., Gamio, J., Bukhari, S., and Yang, W. (2005). Tomographyfor multi-phase flow measurement in the oil industry. FlowMeasurement and Instrumentation, 16(2-3), 145–155.

Johansen, G., Frøystein, T., Hjertaker, B.T., and Olsen, O. (1996).A dual sensor flow imaging tomographic system. MeasurementScience and Technology, 7(3), 297–307.

Jorin (2014). Jorin - Vipa product range. URL http://www.jorin.

co.uk/products/.Kharoua, N., Khezzar, L., and Nemouchi, Z. (2010). Hydrocyclones

for De-oiling Applications—A Review. Petroleum Science andTechnology, 28(7), 738–755.

Michael Lindner (2014). Oil condition monitoring usingelectrical conductivity, oelcheck gmbh. URL http://www.

machinerylubrication.com/Read/29407/oil-condition-

monitoring.Miljoestyrelsen (2010). Status for den danske offshorehandlingsplan

til udgangen af 2009. Miljoestyrelsen.NPL (2014). Physical properties of sea water. URL http://www.

kayelaby.npl.co.uk/general_physics/2_7/2_7_9.html.Pawar, S., Murugavel, P., and Lal, D. (2009). Effect of relative

humidity and sea level pressure on electrical conductivity of airover indian ocean. Journal of Geophysical Research: Atmospheres(1984–2012), 114(D2).

Shepard, D. (1968). A two-dimensional interpolation function forirregularly-spaced data. In Proceedings of the 1968 23rd ACMNational Conference, ACM ’68, 517–524. ACM, New York, NY,USA.

Thorn, R., Johansen, G., and Hammer, E. (1999). Three-Phase FlowMeasurement in the Offshore Oil Industry: Is There a Place forProcess Tomography? Industrial Process Tomography, 228–235.

Lukaszyk, S. (2004). A new concept of probability metric and its ap-plications in approximation of scattered data sets. ComputationalMechanics, 33(4), 299–304.

Williams, R.A., Jia, X., West, R.M., Wang, M., Cullivan, J., Bond,J., Faulks, I., Dyakowski, T., Wang, S., Climpson, N., Kostuch,J., and Payton, D. (1999). Industrial monitoring of hydrocycloneoperation using electrical resistance tomography. Minerals Engi-neering, 12(10), 1245–1252.

Wolbert, D., Ma, B.F., Aurelle, Y., and Seureau, J. (1995). Effi-ciency estimation of liquid-liquid Hydrocyclones using trajectoryanalysis. AIChE Journal, 41(6), 1395–1402.

Young, G. and Wakley, W. (1994). Oil-water separation usinghydrocyclones: An experimental search for optimum dimensions.Journal of Petroleum Science and Engineering, 11(93), 37–50.

